Analysis
To ensure that data are analysed correctly and that the results can be published, it is useful to consult a statistician during the planning stage. Early preparation is essential for developing an appropriate study design, determining a sufficient number of study participants and selecting outcome measures that can answer the research question.
- Plan analysis and data management at an early stage
- Align the research question, study design and analysis
- Studies with a confirmatory purpose
- Studies with an exploratory purpose
- Check study data before analysis
- Analyse the study data
- Specific requirements for medical devices
- Analysing data from a clinical investigation of a medical device
- Analysing data from a performance study of an in vitro diagnostic medical device
- Related information about Advanced therapy medicinal products (ATMPs)
Plan analysis and data management at an early stage
Analysis and data management need to be planned before a clinical study begins.
The planning may be brief and included in the research plan, or it may take the form of a separate document, such as a more detailed Statistical Analysis Plan (SAP).
The more confirmatory the purpose of the study, the more detailed the planning should be. Exploratory studies may be described in less detail.
Align the research question, study design and analysis
For the study to answer the research question, the following elements need to be aligned and presented as a coherent whole in the research plan, protocol or analysis plan:
- research question or hypothesis
- outcome measures
- study design
- methods of analysis.
It must be clear that what is being measured can be used to answer the research question.
Where relevant, the following analyses should be described in advance:
- interim analyses
- adjustments for confounders
- subgroup analyses.
The level of detail should reflect the purpose of the study.
Studies with a confirmatory purpose
A study with a confirmatory purpose is based on a hypothesis formulated before the study begins. Once the study has ended, it must be possible to test the hypothesis using the collected data and the methods described in the research plan, protocol or analysis plan.
It is particularly important that:
- the primary outcome measure is clearly defined
- the sample size is calculated based on the primary outcome measure
- the same method used to calculate the sample size is used for hypothesis testing.
To reduce the risk of bias, the study should, where possible:
- include a control group receiving a placebo or being examined using an established reference method, known as a gold standard
- use randomisation
- be double-blind.
Double-blinding means that neither the study participants nor the researchers know which treatment has been administered.
Studies with an exploratory purpose
xploratory studies aim to generate new knowledge where previous evidence is lacking. Hypothesis testing is not required in these studies.
This often means that:
- the sample size does not normally need to be calculated
- a primary outcome measure does not need to be defined
- control groups, blinding and randomisation are often absent.
The results are mainly reported using descriptive measures, sometimes supplemented by confidence intervals and statistical tests.
The aim is to generate hypotheses for future studies with a confirmatory purpose.
A study with a confirmatory purpose may also include an exploratory component.
Check study data before analysis
Once the practical conduct of the study has ended and all data have been collected in a database or data file, the data need to be checked before analysis.
Check that:
- the data are complete
- any errors or ambiguities are identified and addressed.
All corrections must be documented and retained
Manage corrections in blinded studies
If the study is blinded, corrections must be made without breaking the blind. This reduces the risk of data being corrected or removed on inappropriate grounds.
If the analysis plan needs to be amended, this should be done before the blind is broken. The reasons for the amendment must be documented and clearly reported when the study results are reported and published.
Amendments to the analysis plan may, in some cases, be considered substantial amendments requiring approval from the Swedish Ethical Review Authority.
If the study is not blinded, an independent party may correct and check the data.
Close or lock the study data
Once data checks and corrections have been completed, the study data must be closed or locked. This means that no further changes may be made to the raw data.
Database lock is particularly important in confirmatory studies but may also be used in exploratory studies.
If a database with a locking function is not available, it is important to ensure that an exact copy of the raw data file is preserved unchanged.
Archive study data
The raw data file or database should be archived for possible future review and verification.
Analyse the study data
Before the study’s research questions are analysed, the data may need to be processed.
This may, for example, involve:
- constructing variables from collected data, such as using weight and height to calculate BMI
- replacing missing values through imputation.
If imputation methods are used, they should be described in advance in the research plan, protocol or analysis plan.
The statistical methods used must be based on the research question and study design as described in the research plan, protocol or analysis plan.
If the study has a primary research question, the analysis of the primary variable must be clearly distinguished from analyses of the other variables. The same applies when the study results are presented.
Some common aspects to consider during the analysis stage are described below.
Multiplicity / Multiple comparisons
If several treatments or interventions are analysed against the same outcome measure, the issue of multiple comparisons should be considered. This may mean that the significance level needs to be adjusted.
Confounders
In studies without randomisation, it is important to identify and define in advance any confounders for which the main analyses should be adjusted.
A confounder is a variable that affects both the primary outcome measure and the treatment groups in the study.
Confounders can be identified in different ways, for example by using Directed Acyclic Graphs (DAGs).
In randomised studies, post hoc adjustments for confounders are not normally required.
Subsidiary group analyses and interactions
If an outcome measure covaries or interacts with a variable other than the study treatment, it may be relevant to report the interaction.
The interaction variable may, for example, be:
- a variable that divides the study participants into subgroups, such as sex or age group
- a continuous variable, such as age or weight.
The variables and methods to be used for interaction analyses should, as far as possible, be specified in the analysis plan.
Subgroup analyses may also be exploratory, for example if the analysis reveals previously unknown interactions.
Interim analyses
Interim analyses are conducted while the study is ongoing. Their purpose is to provide a basis for deciding whether the study should continue or be terminated early.
Because interim analyses affect the study’s sample size, it is important that they are carefully planned and described in the research plan or protocol.
Interim analyses are usually conducted by the study’s Data Safety and Monitoring Board (DSMB), an independent review group.
If the blind is broken, the information must remain within the independent group. All other study staff must remain blinded to the results.
Unplanned interim analyses should be avoided because they may reduce confidence in the study’s conclusions.
Points to consider during analysis
When analysing the study results, consider whether the research question has been answered and how the findings should be interpreted.
- Were the data analysed according to the analysis plan?
- Has the original research question been answered?
- How reliable are the results, and what limitations should be considered?
- What conclusions can be drawn, and what are their implications?
- Have the results raised any new research questions?
Specific requirements for medical devices
Specific requirements apply to the analysis of data from clinical investigations of medical devices and performance studies of in vitro diagnostic medical devices.
Contact your regional node within Clinical Studies Sweden if you need statistical support or guidance during the analysis of your study.
Analysing data from a clinical investigation of a medical device
Clinical investigations of medical devices are governed by the EU Medical Devices Regulation (EU) 2017/745 (MDR), which came into effect on 26 May 2021.
Principles of data management and statistical analysis
ISO 14155:2026 applies to clinical investigations of medical devices. Among other requirements, the Clinical Investigation Plan (CIP) must describe the key statistical considerations for the investigation, as outlined in Annex A of the standard.
Statistical planning must be informed by the risk-management process and the clinical evaluation. Together, these processes help determine which clinical data are needed to demonstrate the safety and performance of the medical device.
More information about risk management and clinical evaluation is available in the Idea chapter.
ISO 14155:2026, Swedish Institute for Standards website External link.
Assess changes to the statistical plan
Any deviation from the Clinical Investigation Plan that affects the statistical aspects of the investigation must be assessed to determine whether it constitutes a substantial modification.
A substantial modification must be approved by both the Swedish Ethical Review Authority and the Swedish Medical Products Agency before it is implemented.
Plan the comparison
Confirmatory clinical investigations assessing the safety and performance of medical devices are often designed to compare the investigational device with an established method or treatment. The purpose may be to demonstrate that the device performs at least as well as, or better than, the current standard.
Ideally, confirmatory investigations should be randomised and blinded to minimise the risk of bias. However, these methods can be difficult to apply in clinical investigations of medical devices. For example:
- it may not be possible to develop an identical placebo or comparator device
- using a prospective control group may be inappropriate for ethical or practical reasons
The appropriate approach also depends on whether the medical device has a therapeutic or diagnostic purpose.
Medical devices with a therapeutic purpose
Control group
Where feasible, a prospective control group receiving standard treatment is preferred.
If a prospective control group cannot be used, a historical control may be considered. This means that outcomes from previously treated or untreated populations are compared with outcomes from the participants in the clinical investigation.
Randomisation and blinding
When a historical control is used, randomisation and blinding are not possible.
Even when the investigation includes a prospective control group, it may be difficult to blind participants and investigators if a suitable placebo device cannot be developed.
In these cases, an independent evaluator may assess the outcomes in a blinded manner. The evaluator should not know which treatment each participant received.
Medical devices with a diagnostic purpose
Comparator
In diagnostic investigations, an established method for making the diagnosis is often available. The known diagnostic performance of this method, including its sensitivity and specificity, can be used as a comparator when evaluating the investigational device.
Randomisation and blinding
In some diagnostic investigations, participants may serve as their own controls by undergoing examinations using both the established diagnostic method and the investigational device.
To protect participants, the initial diagnosis is made using the established method. An independent physician may then review the results from the established and investigational methods in a blinded and randomised manner to assess the agreement between them.
Consult a statistician
As in all clinical investigations, the study design, data management and statistical analysis should be reviewed in consultation with a statistician.
The regional nodes within Clinical Studies Sweden provide statistical support during different stages of the research process. Contact your regional node to find out what support is available for your clinical investigation.
ICH E9, Statistical Principles for Clinical Trials, was primarily developed for clinical trials of medicinal products. However, its statistical principles can also be applied to clinical investigations of medical devices.
Medical devices for in vitro diagnostics (IVDR)
Analysing data from a performance study of an in vitro diagnostic medical device
Performance studies of in vitro diagnostic medical devices are governed by the EU In Vitro Diagnostic Medical Devices Regulation (EU) 2017/746 (IVDR), which came into effect on 26 May 2022.
Principles of data management and analysis
ISO 20916:2024 on good study practice applies to clinical performance studies of in vitro diagnostic medical devices.
Among other requirements, the Clinical Performance Study Plan (CPSP) must describe the relevant statistical considerations for the study.
Performance studies are conducted to evaluate the analytical or clinical performance of an IVD device:
- Analytical performance is the ability of a device to correctly detect or measure a particular analyte.
- Clinical performance is the ability of a device to produce results that correlate with a particular clinical condition or with a physiological or pathological process or state in the target population and for the intended user.
ISO 20916:2024, Swedish Institute for Standards website External link.
Evaluate clinical performance
To evaluate the clinical performance of an IVD device, the medical information obtained using the device may be compared with medical information obtained using other diagnostic methods.
More information about comparisons in diagnostic investigations is available in the section Medical devices with a diagnostic purpose.
Consult a statistician
As in all clinical studies, the study design, data management and statistical analysis should be discussed with a statistician when needed.
The regional nodes within Clinical Studies Sweden provide statistical support during different stages of the research process. Contact your regional node to find out what support is available for your performance study.
Related information about Advanced therapy medicinal products (ATMPs)
Advanced therapy medicinal products (ATMPs)
Advanced therapy medicinal products (ATMPs)
Advanced therapy medicinal products (ATMPs) are biological medicinal products based on genes, cells or tissue engineering. ATMPs are regulated by common EU legislation supplemented by Swedish legislation.
The links below provide further information, regulatory guidance, templates and training.
- Advanced therapy medicinal products, Swedish Medical Products Agency website External link.
- Guides, templates and regulatory support, ATMP Sweden website External link.
- ATMP classification and regulatory information, European Medicines Agency website External link.
- Online course: Introduction to advanced therapies, Swedish Academy of Pharmaceutical Sciences website External link.
Research support for clinical trials involving advanced therapy medicinal products
Contact your regional node within Clinical Studies Sweden to find out what research support is available for your study.
Contact your regional node within Clinical Studies Sweden External link.
Published: